on the structure of the information space (an informal ...composing utterances utterances: chains of...
TRANSCRIPT
On the Structure of the Information Space
(an informal overview)
- presented at ER 2009 -
A. L. Furtado, M. A. Casanova,
K. K. Breitman, S. D. J. Barbosa
Departamento de Informática
Pontifícia Universidade Católica do Rio de Janeiro
People make history
• Charles W. Bachman
“the programmer as navigator” – IDS,
CODASYL, pointers, machines, files
• Edgar F. Codd
“the casual user” –
logic, algebra, tables as abstract data types
• Peter P. Chen –
“modeling the things in the real world” –
diagrams, pictograms, concepts
Designing
along 3 stages
Conceptual stage – Entity-Relationship model
Logical stage – Abstract data type
Physical stage – DBMSs
From database design to full-
fledged information system design
“Active conceptual modeling” (Chen, ER 2006).
Our proposal: at the conceptual stage, extend
the ER approach to treat, not only facts, but
also events and agents.
Approach: notions taken from Semiotics
Scope: application domains including public
and business administration, literary genres,
education and training ...
Topics
Part 1 – Design at the conceptual stage
• Facts
• Events
• Agents
Part 2 – Design at the logical stage
• Facts
• Events
• Agents
Part 3 – Example application: a literary genre
Concluding remarks
Part 1
Design at the
Conceptual stage
Three-schema
ER conceptual specifications
static schema – facts – entities, relationships,
attributes, is-a, part-of, ...
dynamic schema – events – application-domain
operations (STRIPS), plans, ...
behavioural schema – agents – situation-goal
rules, typical plans, ...
Conceptual stage:
Facts
Facts as elements of the
Information Space
The ER model: entities and their properties
(attributes and binary relationships)
Facts – assertions about existing
entity instances and their properties
State – all facts holding at a moment of time
Situation – a logical expression involving facts
Composing utterances
Utterances: chains of facts
Saussurean model
syntagmatic axis: composing the chain –Joe‟s age is 25 and Joe works for Acme and…
paradigmatic axis: selecting alternatives
for certain positions in the chain –
Joe or Moe or…
age 25 or age 38 or age 7 or…
Differences within a paradigm
Joe‟s age is 25 and Joe works for Acme and…
Moe‟s age is 38 and Moe works for Acme and…
(*) Joe‟s age is 7 and Joe works for Acme and…
Differences may or may not be “functional”
The axes are not orthogonal:
– integrity constraints, business rules, conventions,…
Conflicts, binary oppositions, negation
On the choice of paradigms:
classes and classification
Property irregularities:
– unknown, non-applicable, defaults, diversity.
Arbitrariness of pre-defined classes.
Lakoff‟s claim: construct classes around typical
representatives, similarity indicators.
Use clustering methods.
From standard to pragmatic time-varying classes
(“all I need to know for my trip”), folksonomies.
Similarity and analogy
Similarity: in the same domain.
Analogy: across different domains.
Using analogy to construct new classes.
Fauconnier and Turner‟s four-space approach:
source, target, generic, and blend
(employee, student, person, trainee)
Map the analogous properties, creative conflict
resolution: re-use and adapt.
Fauconnier and Turner‟s
four-space approach
Going down to details
Semantic hierarchies (modularization): is-a, part-of,…
The Product Division is part-of Acme
The Sales Division is part-of Acme
Joe is assigned to the Sales Division and Moe is assigned to the Sales Division and Moe reports to Joe
Already observed by Saussure – several successive
paradigmatic / syntagmatic planes (structure-preserving mappings):
sentences, words, morphemes, …
Six kinds of part-of
Component / Integral Object - handle / cup
Member / Collection – card / deck
Portion / Mass – slice / pie
Stuff / Object – gin / martini
Feature / Activity – paying / shopping
Place / Area – Everglades / Florida
Characteristics of utterances
Coherent, cohesive: e.g. constituent facts about the
same entity, plus navigation across links:
Joe is assigned to the Sales Division and Moe is assigned to the
Sales Division and Moe reports to Joe
It may be possible to select among alternatives
But the composition is restricted by integrity
constraints and other rules
Descriptions at different levels of detail
Relations between facts
(consequence of the specification)
Syntagmatic – coherence inside an
utterance
Paradigmatic – alternatives within a
common paradigm
Antithetic – negative restrictions imposed
on the information space
Meronymic – successive levels of detail
(semantic hierarchies)
Structure of the Information Space
- an intuitive view -
• •
•
syntagmatic
paradigmatic
antithetic
meronymic
The four master tropes(Kenneth Burke, 1969; Hayden White,1973)
Metonymy – contiguity, relatedness through direct association
→ syntagmatic
Metaphor – similarity despite difference
→ paradigmatic
Irony – marked direct opposition
→ antithetic
Synecdoche – relatedness through categorical hierarchy
→ meronymic
“They are the basic rhetorical structures by which we make sense of experience.”(Jonathan Culler, 2009)
Expect the unexpected
“Marked” states will (ironically) arise!
Wrong beliefs concerning facts and rules,
misconceptions, miscontruals.
Cooperative responses involving data and
metadata.
Double-loop learning, deconstructing (Derrida,
Culler) the design.
Leave room for error, fraud, contradiction, and
exceptional situations.
Conceptual stage:
Events
Modelling events
plot = partially-ordered sequence of events
events = associated with operations executed
by agents, defined by pre-/post-conditions
plots = plans (obtained by a plan-generator)
(not all is covered: non-determinism, natural
events, external agents, ...)
4-sided view of composition process - results from:
4 relations between events (same as for facts)
Plots and Saussure‟s axes
Saussure‟s work in linguistics:
syntagmatic and paradigmatic axes
two dimensions (not orthogonal!)
• syntagmatic: positions in the plot (horizontal axis)
• paradigmatic: choices for positions (vertical axis)
which events can be in some position in a plot?
answer: the events must be related somehow:
• horizontal sequence – syntagmatic relation
• vertical choice – paradigmatic relation
Some “normal” plots
syntagmatic relation event1 and event2:
if event1 leaves the world in a situation that enables the
occurrence of event2 – example:
abduct followed by rescue
paradigmatic relation event1a or event1b:
if event1a and event1b produce a similar effect on the world
– example:
abduct rescueor or
elope capture
Some transgressive plots
antithetic relation if the occurrence of each of
two events presupposes contexts that are (in
principle...) incompatible – examples:
abduct followed by capture
(unnecessary use of force, possibly wrong belief)
elope followed by rescue
(different love feelings)
But suppose there occurs a change concerning beliefs or
even facts...
Zooming in
meronymic relation mapping event1 into event1(i):
if a plan sequence involving event1(i) , for i=1..n, gives a lower level rendering of event1 – example:
abduct can be unfolded into:
ride seize carry
Remark: the 2 first relations induce a 2-dimensional space,
crossed in an oblique angle by the antithetic relation. The
meronymic relation introduces a third dimension, thus
spanning another 2-dimensional surface wherein the other
three relations recur.
Syntagmatic relations
(diagram)
abduct rescue
captureelope
Paradigmatic relations
(diagram)
abduct rescue
captureelope
Antithetic relations
(diagram)
abduct rescue
captureelope
Meronymic relations (1)
(diagram)
abduct
capture
ride defeat seize carry
Meronymic relations (2)
(diagram)
rescue
elope
ride defeat entreat carry
Plots and the four master tropes
rhetorical figures (Lakoff, Burke, Hayden White):
metonymy syntagmatic relation coherence
metaphor paradigmatic relation alternatives
irony antithetic relation sudden shifts
synecdoche meronymic relation details
The external deus ex machina
irony involves extreme binary oppositions:
good/evil, love/hate, strong/weak, etc.
facts: C1 is strong, or
beliefs: C2 believes that C1 is strong
variations in the context, affecting beliefs or facts, allow
unexpected turns in a plot
Aristotle: complex plots feature:
anagnorisis (recognition) and peripeteia (reversal)
Greek theater: god lowered onto the stage using a crane
– our approach: user, through a computer input device...
Dispute for a princess(A brief survey of stories from different countries)
The Sanskrit Ramayana [abduct-rescue]
The Irish Story of Deirdre [elope-capture]
The true case of Patricia Hearst [abduct-capture]
(Stockholm syndrome - Nils Bejerot)
The Roman Rape of the Sabines [abduct-capture]
The Greek legend of Helen of Troy [elope-rescue]
The Tristan and Isolde romance [elope-rescue]
Conceptual stage:
Agents
Modelling agents
Situation-goal rules
Typical plans
Agent profiles – cognitive and affective traits
Roles – buyer, seller, etc.
Roles in folklore genres (Propp):
hero, princess, donor, helper, villain, false hero,
dispatcher
Goal and plan interferences
Robert Willensky - Planning and Understanding - a Computational
Approach to Human Reasoning. Addison-Wesley
(1983).
negative positive
internal conflict overlapping
external competition concord
Classification:
Agents and the four relations
syntagmatic relation - if one favours the other, so that they would
be willing to pursue a joint line of action;
paradigmatic relation - if one is similar to the other, in which case
they can either act independently or seek to emulate each other in
the quest for some goal;
antithetic relation - if one opposes the other, in which case they
behave as enemies;
meronymic relation - if one is an individual and the other is either a
hierarchical superior or some group or organization of which the
former is part (e.g. a troop of soldiers, the inhabitants of a town,
the members of a knightly fellowship, etc.).
Human (as opposed to machine)
decision-making
He [the English philosopher Herbert Spencer] made
parallel lists of reasons for and against the move,
giving each reason a numerical value. The sums
being 110 points for remaining in England and 301
for going [to New Zealand], he remained
(Will Durant, The Story of Philosophy).
“”
You strive and strive,
but what do you seek?
Li T‟ai-Po
Drives, attitudes, emotions, beliefs
Situations motivate goals - trigger situation-goal rules.
A specific goal is just one way to satisfy one or more upper-level
goals – e.g. “raise price” < “increase profit”.
Decide what to do (goals) - drives at the top of goal hierarchies:
sense of duty, material gain, pleasure seeking, spiritual endeavour.
Decide how to do (plans) - attitudes: pleasing, adaptable, outgoing,
careful, self-controlled.
Decide whether or not to commit: emotional satisfaction expected
at goal state, as compared to the current state - anger, disgust,
fear, joy, sorrow, surprise.
“To believe” (rightly or not) rather than “to know”.
Part 2
Design at the
Logical stage
Logical stage:
Facts
To represent and handle facts
---- corresponding to the static schema:
an abstract data type:
frames
frame-sets
frame-manipulation algebra (FMA)
Design at the logical stage:
From tables to frames and frame-sets
In the World Wide Web environment, data comes from multiple
sources, on a highly irregular basis.
Whereas relational tables are homogeneous (nulls are
exceptions), must be in first normal form, and union compatibility is
required for certain operations - but these restrictions are not
inherent in the ER model!
Frames, with a long tradition in AI applications, provides a more
flexible ER-compatible abstract data type for passing from the
conceptual to the logical stage.
In turn, frames and frame-sets can be conveniently converted into
RDF representation at the physical design stage.
Frames and frame-sets - examples
Class employee:[name:_, age:_, salary:_, works/1:_]
Class works: [name:_,cname:_,status:_]
Mary: [name:'Mary', salary:150,
works/1:'Acme']
Acme: [cname:'Acme', headquarters:'Carfax',
works/2:['John','Mary']]
Acme employees: [[name:'Mary', salary:150,
works/1:'Acme'],
[name: 'John', age: 46,
salary: 100, scholarship: 50,
works/1: 'Acme']]
A semiotic view of „completeness‟ (1)
Taking the LISP primitives as example
List – a single data structure for chains and sets
Composing a list: CONS
Extracting from the list: CAR, CDR
For chains, where only the positions matter, this is enough
For sets, it is necessary to extract by comparing values: EQ
Negation: NOT
A semiotic view of „completeness‟ (2)
The LISP primitives cover the first three relations between
facts:
Syntagmatic: compose a chain – CONS
extract from chain – CAR, CDR
Paradigmatic: collect in a set – CONS,
select from set – CAR, CDR, EQ
Antithetic: NOT
A semiotic view of „completeness‟ (3)
The basic Relational Algebra operators (over first-normal-form tables) also
cover the first three relations between facts:
Syntagmatic – product, projection
Paradigmatic – union, selection
Antithetic – difference
Remark:
completeness proved through a comparison with Relational Calculus – but
semiotic completeness can also be claimed ……
…… except that NF2 tables would need additional operators
Frame Manipulation Algebra (FMA)
operations
Defined on frames and frame-sets
Executable as embedded in a logic programming language
Unification and most specific generalization over frames and frame-
patterns are also provided
Operations:
Syntagmatic – product, projection
Paradigmatic – union, selection
Antithetic – difference
Meronymic – combination, factoring
Logical stage:
Events
To represent and handle events
---- corresponding to the dynamic schema:
an abstract data type:
plots (which are frame-like structures)
libraries (sets of plots)
plot manipulation algebra (PMA)
An engine:
Plan-generation / Plan-recognition
Plan-generation:
executable specifications, simulation, online
access to conceptual schema
Plan-recognition:
typical plans, re-use, check what a person is
trying to do, logs and plot mining
Logical stage:
Agents
To represent and handle agents
---- corresponding to the behavioural schema:
ongoing research:
■ drawing from Elaine Rich‟s work using
frame-like stereotypes, to represent cognitive
and affective chacteristics of agents
Part 3
Example application:
a literary genre
Swords and Dragons:
ER diagram
hero victim donor
Creature
name
alive
gender
Dragon Person
Knight Magician Princess
current place
kidnapped
affection
level
protectn
place name home
acquaintance
married
villain
nature
strength
kind
Place
Swords and Dragons:
Hierarchy of typical plans
end
adventure
avengerescue
retaliatedo_villany donate accomp.
liberate help false_helpexecutemurderabduct
kill free marryfightkidnapreduce
protection
attack
Swords and Dragons:
Static schema
entity(character,name).
entity(person,name).
entity(knight,name).
entity(princess,name).
….
is_a(knight,person).
is_a(princess,person).
is_a(magician,person).
is_a(dragon,character).
role(hero,knight).
….
attribute(character,strength).
relationship(home,[character,place]).
relationship(current_place,[character,place]).
relationship(acquaintance,[character,character]).
….
Swords and Dragons:
Dynamic schema….
operator(5,
fight(CH1,CH2),
[ alive(CH1), alive(CH2),
nature(CH1,KIND1),
nature(CH2,KIND2),
dif(KIND1,KIND2),
dif(KIND1,0.0), dif(KIND2,0.0),
strength(CH1,LS1), strength(CH2,LS2),
{LS1>=10.0, LS2>=10.0},
current_place(CH2,PL), current_place(CH1,PL),
protection(PL,[KIND3,L_PROT]),
{L_PROT=<0.0,
NEW_LS1=LS1-LS2,
NEW_LS2=LS2-LS1} ],
[not(strength(CH1,LS1)), not(strength(CH2,LS2)),
strength(CH1,NEW_LS1), strength(CH2,NEW_LS2)],
10,
[strength(CH1,NEW_LS1), strength(CH2,NEW_LS2)],
[],[]):-
db(character(CH1)),
db(character(CH2)).
….
Swords and Dragons:
Behavioural schema
….
/* The strongest hero wants to become stronger
than the villain */
rule([ e(i,strength(HERO,Lh)),
e(i,villain(VIL)),
e(i,strength(VIL,Lv)),
h({Lh=<Lv}) ],
([T],
[ h(T,strength(HERO,LS)),
h({LS > Lv}),
h(T>i)],
true))
:- findall(S,(db(strength(H,S)),db(hero(H))),Ss),
max_list(Ss,Lh),
db(hero(HERO)),
db(strength(HERO,Lh)).
….
The Logtell prototype – plot composition
The Logtell prototype – animation
Concluding remarks
The importance of Semiotics to the design of
Information Systems
The importance of a Computer Modelling
approach to Semiotics
Main points of our project, at this moment:
• the four semiotic relations
• frames (and frame-like structures) at the logical stage
• executable specifications – prototype tools
Project bibliography
A.E.M. Ciarlini, M.A. Casanova, A.L. Furtado, P.A.S. Veloso: “Modeling Interactive Storytelling
Genres as Application Domains”. Journal of Intelligent Information Systems (to appear).
A.E.M. Ciarlini, S.D.J. Barbosa, M.A. Casanova, A.L. Furtado: “Event Relations in Plan-Based
Plot Composition”. ACM Computers in Entertainment 2009: 4(to appear).
A.L. Furtado, M.A. Casanova, K.K. Breitman, S.D.J. Barbosa: “A Frame Manipulation Algebra
for ER Logical Stage Modelling”. ER 2009.
A.L. Furtado: “A Decision-Making Process for Digital Storytelling”. MCC 21 2009
B. Karlsson, S.D.J. Barbosa, A.L. Furtado, M.A. Casanova: “A Plot-Manipulation Algebra to
Support Digital Storytelling”. ICEC 2009: 132-144.
A.L. Furtado, M.A. Casanova, S.D.J. Barbosa, K.K. Breitman: “Analysis and Reuse of Plots
Using Similarity and Analogy”. ER 2008: 355-368.
M.A. Casanova, S.D.J. Barbosa, K.K. Breitman, A.L. Furtado: “Generalization and Blending in
the Generation of Entity-Relationship Schemas by Analogy”. ICEIS 2008.
K.K., Breitman, S.D.J. Barbosa, M.A. Casanova, A.L. Furtado: “Conceptual modeling by
analogy and metaphor”. CIKM 2007: 865-868.
K.K. Breitman, S.D.J. Barbosa, M.A. Casanova, A.L. Furtado, M. G. Hinchey: “Using Analogy to
Promote Conceptual Modeling Reuse”. ISoLA 2007: 111-122.
S.D.J. Barbosa, K.K. Breitman, A.L. Furtado, M.A. Casanova: “Similarity and Analogy over
Application Domains”. SBBD 2007: 238-252.
A.E.M. Ciarlini, C.T. Pozzer, A.L. Furtado, B. Feijó: A logic-based tool for interactive generation
and dramatization of stories. Advances in Computer Entertainment Technology 2005: 133-140.